Automated Framework to Translate Rebar Spatial Information from GPR into BIM

Abstract

Automated three-dimensional (3D) as-built modeling of built infrastructure with embedded rebar is an ongoing area of research. The current practice for such a process is either highly manual, due to the need for localizing rebar during the scanning process, or not compatible with popular and widely used building information modeling (BIM) software. In order to fill the existing gap, this paper proposes a highly integrated approach that utilizes ground-penetrating radar (GPR) to detect rebar from in-service buildings and automatically translate the extracted rebar from GPR data into existing BIM. Rebar is localized through GPR data using the shearlet transform and the inverse correlation between the maximum intensity and depth parameters. To map the localized rebar into corresponding building elements in BIM, a classification system is developed based on a deep-learning algorithm trained by several on-site data sets. To evaluate the performance of the presented system, two buildings and their various concrete components are selected as the testbed. The obtained results are promising and reveal that (1) the proposed GPR2BIM system is highly automated and accurately integrates localized rebar into existing BIM; (2) the shearlet transform precisely recognizes rebar through GPR data; (3) the lowest absolute error for the depth estimation algorithm is 5.12%; and (4) rebar from GPR is automatically classified and assigned to corresponding building elements with an accuracy of 100%. The findings demonstrate that the highly integrated system developed in this research is capable of automatically building a complete BIM with both surface elements and rebar for in-service building.